Identification of a PTEN-regulated STAT3 brain tumor suppressor pathway

Abstract

Activation of the transcription factor STAT3 is thought to potently promote oncogenesis in a variety of tissues, leading to intense efforts to develop STAT3 inhibitors for many tumors, including the highly malignant brain tumor glioblastoma. However, the function of STAT3 in glioblastoma pathogenesis has remained unknown. Here, we report that STAT3 plays a pro-oncogenic or tumor-suppressive role depending on the mutational profile of the tumor. Deficiency of the tumor suppressor PTEN triggers a cascade that inhibits STAT3 signaling in murine astrocytes and human glioblastoma tumors. Specifically, we forge a direct link between the PTEN-Akt-FOXO axis and the leukemia inhibitory factor receptor beta (LIFRbeta)-STAT3 signaling pathway. Accordingly, PTEN knockdown induces efficient malignant transformation of astrocytes upon knockout of the STAT3 gene. Remarkably, in contrast to the tumor-suppressive function of STAT3 in the PTEN pathway, STAT3 forms a complex with the oncoprotein epidermal growth factor receptor type III variant (EGFRvIII) in the nucleus and thereby mediates EGFRvIII-induced glial transformation. These findings indicate that STAT3 plays opposing roles in glial transformation depending on the genetic background of the tumor, providing the rationale for tailored therapeutic intervention in glioblastoma.

Figure 1.

PTEN deficiency suppresses the LIFRβ–STAT3 signaling pathway. (A) Lysates of mouse Pten^wt/wt or Pten^−/− astrocytes were immunoblotted with antibodies to LIFRβ, Tyr705-phosphorylated STAT3 (pYS3), total STAT3, PTEN, total Akt, Ser473-phosphorylated Akt (pAkt), or actin. (B) Immunoblotting of parental Pten^loxP/loxP and Pten^−/− astrocytes with antibodies used in A. Loss of PTEN expression was associated with an increase in phospho-Akt levels and with a decrease in LIFRβ and pYS3. Actin served as loading control. (C, left panel) Immunoblotting of lysates of Pten^−/− astrocytes treated with the PI3K inhibitor LY294002 or DMSO vehicle control for 48 h. Asterisk indicates a nonspecific band. (Right panel) Immunoblotting of lysates from serum-starved Pten^loxP/loxP astrocytes stably expressing a constitutively active form of Akt (caAkt). Activated Akt reduced LIFRβ and pYS3 levels. (D) Immunocytochemical analysis of Pten^loxP/loxP and Pten^−/− astrocytes with a FOXO3 antibody. Nuclei were stained with a DNA dye (Hoechst). FOXO3 was excluded from the nucleus in Pten^−/− astrocytes. (E) Pten^loxP/loxP and Pten^−/− astrocytes were transfected with a luciferase reporter gene controlled by FOXO-binding sites (FHRE-Luc), together with a renilla expression plasmid to serve as an internal control, and subjected to a dual luciferase assay. FOXO-dependent transcription was reduced in Pten^−/− astrocytes. (F) ChIP analysis at the endogenous LIFRβ promoter in Pten^loxP/loxP and Pten^−/− astrocytes with a FOXO3 antibody. A rabbit anti-HA antibody was used as negative control. The analysis was done with two independent sets of primers for the LIFRβ promoter. Negative controls for the PCR reaction were performed with primers for the E-cadherin promoter. Endogenous FOXO3 occupied the endogenous LIFRβ promoter in Pten^loxP/loxP but not in Pten^−/− astrocytes. (G) Immunoblotting with the LIFRβ and FOXO3 antibodies of Pten^loxP/loxP and Pten^−/− astrocytes infected with a FOXO3 RNAi-encoding lentivirus (FOXOi) or an empty vector and selected with puromycin. Actin served as loading control. Knockdown of endogenous FOXO3 reproducibly led to down-regulation of LIFRβ. (H) Immunoblotting of lysates of serum-starved Pten^−/− astrocytes transfected with a LIFRβ expression plasmid or a control vector that were left untreated (−) or treated with LIF (L) for 15 min. LIFRβ restored LIF-induced STAT3 phosphorylation in Pten^−/− astrocytes.

Figure 2.

STAT3 knockout promotes astrocyte cell proliferation and invasiveness. (A, left panel) Immunoblotting of Stat3^loxP/loxP or Stat3^−/− astrocytes with an antibody to STAT3. Actin served as control for loading. (Right panel) Cell population growth of Stat3^loxP/loxP and Stat3^−/− astrocytes. STAT3 loss significantly increased astrocyte cell population growth rate (representative experiment of three independent experiments performed in triplicate; ANOVA; [*] P < 0.05; [**] P < 0.0001). (B) Quantification of the invasive potential of Stat3^−/− astrocytes through a matrigel substrate. Stat3^loxP/loxP and Stat3^−/− astrocytes were seeded on top of an 8-μm pore size insert coated with matrigel and allowed to invade through the matrigel matrix for 22 h. Knockout of STAT3 significantly increased astrocyte cell invasiveness (n = 3; t-test; [*] P < 0.01). The effect of STAT3 loss on invasiveness was not secondary to a change in cell proliferation, as the invasive potential of these cells was measured at a time (22 h after plating) prior to a significant increase in cell number upon STAT3 knockout. (C) Immunoblotting of STAT3 and PTEN in Stat3^loxP/loxP or Stat3^−/− astrocytes that were uninfected or infected with a retrovirus encoding a shRNA directed against PTEN (shPTEN). Actin served as control for loading. (D) Cell population growth of PTEN knockdown (PTEN^shRNA) astrocytes. STAT3 loss significantly increased cell number of PTEN knockdown astrocytes (representative experiment of two independent experiments performed in triplicate; ANOVA; [*] P < 0.0001).

Figure 3.

STAT3 suppresses PTEN deficiency-induced malignant cell transformation. (A,B) Stat3^loxP/loxP;PTEN^shRNA or Stat3^−/−;PTEN^shRNA astrocytes were injected subcutaneously into SCID mice. Eight weeks after injection, tumors were removed, measured, and stained. STAT3 loss induced an increase in tumor size as revealed by hematoxylin and eosin (H&E) staining (A) and tumor size measurements (n = 8; t-test, [*] P < 0.05) (B). Arrows in A show the tumor limits. Bar, 1 mm. (C) Histologic analysis of the Stat3^−/−;PTEN^shRNA tumors by H&E staining. The tumors showed histologic features of neoplastic transformation including nuclear atypia, pleomorphism, and frequent mitotic figures (arrowheads). Bar, 100 μm. (D) Nestin and Ki67 immunostaining of Stat3^loxP/loxP;PTEN^shRNA and Stat3^−/−;PTEN^shRNA tumors. These tumors express nestin, a characteristic marker of glial tumors. Cell proliferation rate, as measured by the percentage of Ki67-positive cells, was higher in Stat3^−/−;PTEN^shRNA tumors as compared with Stat3^loxP/loxP;PTEN^shRNA tumors (64% vs. 25%, respectively; average of two tumors each). Bar, 100 μm.

Figure 4.

STAT3 interacts with EGFRvIII in the nucleus. (A, left panel) Lysates of 293T cells transfected with expression plasmids encoding Flag-STAT3, EGFRvIII, or both were immunprecipitated with a Flag antibody followed by immunoblotting with an EGFR or STAT3 antibody. Total lysates (Input) were also immunoblotted with these antibodies. Actin served as loading control. EGFRvIII interacted efficiently with STAT3. (Right panel) Lysates of stable MSCV-GFP and MSCV-GFP/EGFRvIII astrocytes were immunoprecipitated with an EGFRvIII antibody followed by immunoblotting with an EGFR or STAT3 antibody. Total lysates (Input) were also immunoblotted with these antibodies. EGFRvIII interacted with endogenous STAT3 in astrocytes. (B) Immunoblotting of Pten^loxP/loxP and Pten^−/− astrocytes, or stable MSCV-GFP and MSCV-GFP/EGFRvIII astrocytes with antibodies that recognize Tyr705-phosphorylated STAT3 (pYS3), Ser727-phosphorylated STAT3 (pSS3), total STAT3, PTEN, Ser473-phosphorylated Akt (pAkt), total Akt, and EGFR. Actin was used as loading control. EGFRvIII expression had little effect on STAT3 Tyr705 and Ser727 phosphorylation. (C) Lysates of 293T cells transfected with Flag-STAT3, EGFRvIII, or both were fractionated into cytoplasmic and nuclear fractions and then subjected to immunoprecipitation and immunoblotting analyses as in A. SnoN and 14–3–3 were used as nuclear or cytosolic markers, respectively. EGFRvIII interacted with STAT3 more efficiently in the nuclear than in the cytoplasmic fraction. (D) EGFRvIII is present in the nucleus of astrocytes. (Top panel) Lysates of EGFRvIII-expressing and control astrocytes were subjected to subcellular fractionation and immunoblotting with the EGFR antibody. Lamin A and α-tubulin were used as nuclear or cytosolic markers, respectively. (Bottom panel) Immortalized astrocytes transfected with a plasmid encoding an EGFRvIII-GFP fusion protein were fixed and imaged by confocal microscopy to detect GFP fluorescence. Hoechst was used to visualize nuclei. Arrowheads indicate the position of the nucleus.

Figure 5.

STAT3 mediates EGFRvIII-induced cell transformation. (A) Immunoblotting of EGFRvIII-expressing or control astrocytes with antibodies against EGFR and STAT3. Actin was used as loading control. (B) Cell population growth curves of Stat3^loxP/loxP; EGFRvIII and Stat3^−/−;EGFRvIII astrocytes. STAT3 knockout significantly decreased cell number (representative experiment of three independent experiments performed in triplicate; ANOVA; [*] P < 0.0001). (C) Size of EGFRvIII-expressing tumors. Stat3^loxP/loxP; EGFRvIII and Stat3^−/−;EGFRvIII astrocytes were injected subcutaneously into SCID mice. Four weeks after injection, tumors were excised, measured, and stained. Tumors were only present in Stat3^loxP/loxP;EGFRvIII-injected mice (n = 6; t-test; [*] P < 0.05). (D) H&E staining of Stat3^loxP/loxP;EGFRvIII tumors confirmed the presence of tumor cells. Mitotic figures are indicated by arrowheads. Bar, 100 μm. (E, left panel) Immunoblotting of PTEN^shRNA;EGFRvIII astrocytes with EGFR and STAT3 antibodies. Actin served as loading control. (Right panel) Size of PTEN^shRNA;EGFRvIII-expressing tumors. PTEN^shRNA;EGFRvIII astrocytes were injected subcutaneously into SCID mice. Four weeks after injection, tumors were excised and measured. Tumors were only present in Stat3^loxP/loxP; PTEN^shRNA;EGFRvIII-injected mice (n = 4; t-test; [*] P < 0.001).

Figure 6.

STAT3 signaling in human glioblastoma specimens. (A) Immunoblotting of lysates of human glioblastoma samples (GBM) with antibodies to EGFR that recognize both wild-type and EGFRvIII forms, PTEN, LIFRβ, Tyr705-phosphorylated STAT3 (pYS3), or total STAT3. GAPDH was used as loading control. (B,C) Spearman correlation matrix of PTEN and EGFRvIII levels (B) or PTEN, LIFRβ, and pYS3 levels (C) measured as continuous variables in the immunoblots shown in A. The Spearman rank correlation coefficients, r_s ([*] P < 0.05; [**] P < 0.005; [***] P < 0.0001), are shown.

Figure 7.

Dual role of STAT3 in tumorigenesis. (A) A PTEN-regulated STAT3 tumor-suppressive pathway. (B) EGFRvIII induces an oncogenic switch in STAT3 function.